NEW · AGENT Security findings. Published fixes. Your fleet in context.

WINDOWS OBSERVABILITY, CONNECTED.

Keep watch.
Stay in control.

An agent that keeps checking for vulnerabilities and published fixes relevant to your fleet. Review the evidence, choose what matters, and carry it into Workbench for your team to act on.

MEET THE AGENT

Always checking.
Ready for your review.

On its schedule, the Agent reads enabled sources and compares them with the software and updates on your hosts. Security and reliability findings arrive with context for an investigation.

CISA KEVNVDMicrosoft updatesEPSSCrowdStrike

Configure sources, priority tags and business-hours scheduling. Follow each run in the activity log and history.

AGENT / FINDINGS TO REVIEWILLUSTRATIVE
SECURITYPublic-facing host

A software update worth prioritising.

Related CVEs grouped by software, with affected hosts and the version that fixes them.

Installed versionVulnerability sourcesFixed version
RELIABILITYPossible fix

A published fix. A matching symptom.

A Microsoft update describes a fix in the same feature area as a recurring service failure. Review the match against the host’s events and installed updates.

01 · Agent finds02 · You review03 · Add to Workbench

Findings wait for a person to review and add them.

WinEvents.io / WorkbenchILLUSTRATIVE DATA

A clear place to start.

Review the signals. Decide the next step.

Last 24 hours
Customer portalFiltered hosts onAI review · just now
Events over time Events Warnings Errors
00:0006:0012:0018:0024:00
Team cardsShared ownership. Visible progress.

Backlog 1

SECURITY

Review an audit policy change

Check the recorded account and confirm the change with the owner.

Security 80/ 100
Alex · 1 comment

In progress 1

PERFORMANCE

Investigate sustained CPU pressure

Compare the busiest process with the IIS response-time window.

Performance 90/ 100
Sam · CHG-1042

Completed 1

RISK

Verify certificate renewal

Review the certificate events and record the server-side check.

✓ Review completed
Jordan · 2 comments
Built on an open data foundation
Cloudflare Workers
R2 SQL
Apache Iceberg
Workers AI

ONE PLATFORM. THE WHOLE PICTURE.

Follow the signal.
Keep the context.

Connect proactive findings with hands-on investigation. Review the Agent’s evidence, track work with your team, and compare the impact of a change.

AGENT · SECURITY & RELIABILITY

Relevant findings.
A human decision.

Match CVEs to the software CrowdStrike sees and the Windows updates each host has installed. Connect crashes, hangs and service failures with Microsoft fixes the host does not yet have.

  • Prioritise public-facing hosts and filter security findings by CVSS
  • Review grouped software findings and suggested investigation paths
  • Add reviewed findings to Workbench, with links that follow their status
AGENT / REVIEW CONTEXTEXAMPLE
Published issue → fleet match → finding
SecuritySoftware, versions & CVEs
ReliabilitySymptoms & published fixes
TimelinePublished · installed · found
Next stepReview, then add to Workbench
Inspect the evidence and history before making a finding part of the team’s work.

AI THAT KEEPS THE EVIDENCE IN VIEW.

A considered review.
Practical next steps.

Workbench combines Workers AI triage with final recommendations and next steps from Claude Sonnet 5.5. Risk, performance and security scores help you decide what to investigate first.

Hostnames, accounts and addresses are replaced with tokens before the review leaves Cloudflare. Reviews are saved for everyone by time range and flagged once they are a day old.

✧Fleet summaryEXAMPLE

LAST 24 HOURS · COMPARED WITH THE DAY BEFORE

A change worth investigating.

WEB-SERVER-02 recorded more 5xx responses during a period of elevated CPU. Review the aligned host timeline to see which events occurred in the same window.

CPU peak94%
5xx requests128
Window14:00–15:00
↳ Correlation gives you somewhere to look. It does not establish the cause.

FROM OVERVIEW TO ANSWERS.

Less switching.
More investigating.

A connected workflow for the questions infrastructure teams ask every day.

01 / ORIENT

See what needs attention.

Start with the Agent’s waiting security and reliability findings. Narrow by host or tag, review the evidence and choose which findings to add to Workbench.

02 / INVESTIGATE

Follow the evidence.

Hover a card to locate its events on the chart, which stays in view as you scroll. Open the underlying records, review the recommendation and add a change request when needed.

03 / FOLLOW THROUGH

Make ownership visible.

Add a card to My cards and move it through Backlog, In progress and Completed. Team cards shows who has what; people, status and comments carry across reviews.

▤

Security, in context.

An expanded security watchlist, failed sign-ins and lockouts, with CrowdStrike host context and Secure Boot certificate status to guide your investigation.

⌁

Web performance, connected.

Explore IIS status codes, response times and failing pages, then open the surrounding Windows events for the same host.

⊞

Software behind the noise.

Trace event volume to its software, inspect application crashes and unusual programs, and see when services stopped or started.

LATEST IN WINEVENTS.IO.

More context.
With every release.

The Agent brings vulnerability and reliability findings into your workflow. New maps connect the evidence, and Computer Data shows where collection volume goes.

v2.46.0LATEST

See what your computers send.

Compare event, metric and IIS data by computer and CrowdStrike site, over time and in a sortable table. Open a host to see which events account for its data.

Release details

Review high-volume event kinds marked low value, kept or worth checking, with explanations and estimated savings. Copy a proposed Vector filter configuration for the selections you choose.

COMPUTER DATA
v2.45.0

Connect recurring problems with published fixes.

The Agent matches 30 days of crashes, hangs, service failures, failed updates and unclean restarts to Microsoft fixes in updates the host does not have.

Release details

Matches use named programs, modules, errors, exceptions and services. Possible fixes in the same feature area are checked by the model and dropped if unlikely. Reliability findings become Risk or Performance cards only when someone adds them to Workbench. Security findings are grouped by software, with a fixed version and a CVSS filter. Finding links follow them through review, dismissal or resolution.

RELIABILITY
v2.44.0

A scheduled agent, grounded in your fleet.

Read enabled vulnerability sources and match them to CrowdStrike application inventory and installed Windows cumulative updates. Prioritise public-facing hosts and review scored findings before adding them to Workbench.

Release details

Sources include CISA KEV, NVD, Microsoft security updates and EPSS, with optional CrowdStrike Spotlight. Configure sources, priority tags and scheduling. Follow published, installed, found and reviewed markers on the timeline, with a history of agent and human activity.

AGENT
v2.43.0

See how the evidence connects.

Explore Account, Dependency, Happens Together, Process and Shared Certificates maps, alongside the Identity Map.

Release details

Use the timeline, refiners, Find and details panels to investigate relationships. Open underlying events in Query and share the map with its selected context.

MAPS
v2.42.0

Take your evidence with you.

Export selected events and IIS requests as CSV, or graph the displayed rows by source, host, severity and other fields.

Release details

Export up to 10,000 selected rows, newest first. Click a graph bar to filter the list. Workbench links can show the latest available review when the chosen scope has none in the current range.

QUERY
v2.32.0

Keep one identity for continuing work.

All cards brings together findings across scopes and ranges, showing ownership, board status and first and last seen.

Release details

Search and sort the list, find unclaimed work and merge duplicate cards while retaining people, comments, change requests and links. Continuing issues reuse their card ID across reviews.

WORKBENCH

Based on the application’s release notes through v2.46.0.

A FEW USEFUL DETAILS.

Built for the
way you investigate.

Does the Agent add work automatically?

The Agent wakes on its configured schedule and produces findings. Nothing reaches Workbench until someone reviews a finding and adds it. From there, your team can take ownership and track the work.

What does the Agent need to match vulnerabilities?

It compares enabled vulnerability sources with application inventory from CrowdStrike and installed Windows cumulative updates. CrowdStrike Spotlight is optional and requires its licence and Vulnerabilities: Read scope. Findings depend on the inventory, update evidence and sources available.

What is a reliability finding?

The Agent compares recent problems in Windows events with published Microsoft fixes in updates a host does not have. A specific match can raise a finding; a possible fix in the same feature area is checked by the model. Review that evidence before deciding whether to apply an update.

What stays with a Workbench card between reviews?

Its people, status and comments carry forward across reviews and time ranges, even when issues are grouped differently. Boards show when a card was last seen and flag open cards that the latest review did not find.

How is data handled in the Workbench AI review?

Workers AI performs triage and Claude Sonnet 5.5 provides final recommendations and next steps. Hostnames, accounts and addresses are replaced by tokens before anything leaves Cloudflare. The “What Claude saw” view lets you inspect what was sent to the final model.

Can I review a change across several servers?

Yes. Select servers by name or tag, mark the change window and compare before, during and after. New changes start visible to you only. You can save for everyone, attach AI analyses and share a link. Saved changes also appear as bands on connected timelines.

Does Certificates list every installed certificate?

The view is based on certificate events reported by Windows, including expiring or expired machine certificates and TLS failures. Use the supplied PowerShell check on the server to confirm a certificate’s name and exact expiry.

What data does WinEvents.io bring together?

Windows event logs, measured host CPU and memory, and IIS requests. The explorer queries data held in R2 Data Catalog as Apache Iceberg tables using R2 SQL.

Can I use it without AI?

Yes. Fleet views, manual search, filters, counts, charts and drill-downs work independently of AI. AI assists with query writing, summaries and explanations when you choose to use it.

How do shared links work?

Links preserve views such as the tab, time range, search and open host. Saved analyses can also be linked when storage is configured. Recipients still need access to the application; a shared link does not grant access.

Does the website show live infrastructure data?

No. The visuals here use illustrative data to explain the product. Open the application to explore your environment with your existing access.

Does Identity show who caused a change?

Identity shows the account recorded by Windows when the event includes one, separately from people who signed in to the host during the preceding hour. A nearby sign-in is context for an investigation, not proof that the person caused the change.

YOUR NEXT INVESTIGATION STARTS HERE.

See the signal.
Carry the work forward.

Prioritised issues. Shared ownership. Connected evidence.